AR256 - SIH 2020
Final Code for the Problem Statement - AR256 by NATIONAL JUTE BOARD,MINISTRY OF TEXTILES.
AI BASED SOLUTION WHICH MAY CALCULATE THE PROBABILITY OF FUTURE MARKET TREND ON THE BASIS OF PAST YEARS DATA AND SUGGEST A SUITABLE PRICING MODEL FOR SALE OF COTTON.
- Observing the trend of cotton prices over the last few years, it can be interpreted that the prices for cotton fiber are highly volatile in our commodity market.
- It has been observed that during the auction for the cotton crops, the buyers have been forming syndicates and deciding on a lesser auction price. This results in a loss for the farmers due to no transparency.
- Most importantly, the farmers need a holistic idea of the market price of cotton in order to make appropriately priced crop-sales.
- Being vigilant about the fluctuations and future trend of prices is necessary for the stakeholders, belonging to the Cotton Sector of the textile industry.
- Our model (Prophet) is already used in industry grade applications for producing reliable forecasts for planning and goal setting.
- We’ve found it to perform much better than any other approach in the prediction of volatile cotton prices.
- Models are trained on Stan so that you get forecasts in just an instant.
- Takes into account Holiday or Harvest season for pricing prediction.
- Model flexible to include user defined seasonality and changepoints.
-
Mean Absolute Error Percentage (MAPE): 13.9% MAPE is the measure of prediction accuracy of a forecasting method in statistics, for example in trend estimation, also used as a loss function for regression problems in machine learning.
-
R2 Score: -0.77 R-Squared Score is a statistic used in the context of statistical models whose main purpose is either the prediction of future outcomes or the testing of hypotheses, on the basis of other related information. Ideal Value of R2 Score= +1 or -1
CottonData.csv is the modified and cleansed data.
model.pckl is the weighted trained model.
Machine Learning, Deep Learning, Facebook Prophet, HTML, CSS, Javascript, Flask, SQLAlchemy
Screen takes were added into the repo
- 14 Unique States
- 163 Unique Districts
- 702 Unique Markets
https://drive.google.com/drive/folders/1VHyg3kw2dyOKr6W2nm_pjcg7dL15dV5j?usp=sharing